Model infrastructureQuality

Task-quality routing

Collect provider outcomes for a named task and use evidence rather than a fixed preference list.

Task-quality routing — contract and behavior

  • The collector runs each active model against the task prompt and records the outcome in the operational journal.
  • Routing selects the highest measured score for the exact task name and randomizes equal scores.
  • Provider metadata comes from the broker while every credential is redeemed independently at invocation time.
  • Task observations are operational evidence, not a secret, but prompts and credentials must not be copied into the journal.

Stay in the loop. Never miss out.

Subscribe to our newsletter and unlock Wisent insights.